PP logo


Power supply soft-starter.



PS37 is a self-contained soft-starter for power supplies.
The circuit can also include a mains-voltage selector.


This circuit ( all of it ) is connected directly to the mains.
Do not work or measure on the circuit while it is powered. It can kill you.
All mains wiring must be done in accordance with your local safety standards.
The design contain NTC Inrush Current Limiters ( ICLs ). These can get very hot ( up to 250 °C ) under fault conditions and must be kept in a safe distance from anything that can melt ( like wire insulation ).
All though the coating on most ICLs is insulating, it is not considered an insulation barrier by safety standards.
ICLs must be installed with the appropriate distance to secondary circuits and enclosure parts.

How not to build a soft-starter.

Simple soft-starter schematic.
Fig.1: Simple soft-starter ( or how not to build a soft-starter ).

Fig.1 shows a very common circuit in the early 1980s with component values as I remember.
The reason it is included here is that some people still think it is a good idea - it is not - it is potentially lethal.
I1 is a mains fuse of typically 5 A to 10 A.
R1 is the soft-start resistor - typically an aluminum cased type around 50 ohms mounted on the chassis.
E1 is a relay that shorts R1 when the voltage on the secondary is high enough for the relay to pull in ( typically 20% to 80% of its rated voltage ).
Fig.1 shows the circuit with an AC relay powered from the primary side of T1, but some designs use a DC relay powered from T1 secondary.
T1 is a transformer of 1-2 kVA, often a toroid to save weight.
Affordable toroids were relatively new on the market then and the very large magnetizing current was a surprise to some people so a soft-start circuit was in some cases added as an after-thought, using chassis mounted components.
The relay has a relatively short life as pulling it in with a slowly rising voltage will cause the contacts to arc during pull in.
In case of a failure on the secondary side ( shorted rectifier or output transistors ), the fuse will blow.
Many people just change the fuse without checking the circuit.
Next time the mains is applied, the relay will not pull in and the full mains voltage is across the resistor dissipating around 1 kW ( 230 V and 50 Ω ). 230 V into 50 Ω is 4.6 A, so a fuse of 5 A or more will not blow - the resistor will.
A failure mode I have seen several times with aluminum cased is that the element shorts to the housing.
Now there is mains voltage on the chassis. Hopefully a good safety-earth connection will cause an over-current or fault-current interrupter to trip.
How ( if ) anybody managed to get this circuit through safety testing is still a mystery to me.
I have checked the datasheet for several aluminum cased resistors. Not one manufacturer suggest that these resistors are designed for or approved for mains voltage.


The following circuit is designed to operate over a mains voltage range from 95 VAC to 135 VAC or 190 VAC to 270 VAC and a temperature range of 0 °C to 70 °C ( 85 °C except for the IC and ICLs ).
The operating temperature should be kept low for maximum component life.

PS37 mains wiring schematic.
Fig.2: PS37 mains wiring.

J1 and J1C is a 6-way pluggable terminal block for connection of mains input and transformer wires.
J2 is a 5-way terminal block header and J2C is a 4-way pluggable terminal block with wires across pins 1-2 and 3-4 for mains voltage selection.
If you do not need a voltage selector, replace J2 with wire-jumpers ( across pins 2-3 and 4-5 for 115 VAC or across pins 3-4 for 230 VAC ).
I1 is a fuse with a universal footprint that will accept a number of 5x20 mm fuse holders and some TR-5 fuse holders.
R1 and R2 are Inrush Current Limiter NTCs. The board will accept ICLs of up to 32 mm diameter.
They will get very hot in case the relay fails and must be mounted with some mm clearance above the PCB.
All components ( terminals, relay, ICLs and fuse-holder ) in this part of the circuit MUST have a current rating so they can blow the fuse.
Fuses will normally blow at a current of 1.5..2 times their rated current. Do read the data-sheet for the fuse type you use.
A 4 A fuse will withstand 7 A for longer time than a 5 A ICL. The ICL will blow - not the fuse.
The connections AC1 and AC2 provides 115 VAC power for the control circuit.

PS37 control circuit schematic.
Fig.3: PS37 control circuit.

R3, R4 and R5 acts as an over-current and over-voltage protection for the circuit.
R3 and R4 are fusible resistors with a peak-voltage rating of 1 kV and a power rating of 0.5 W. These will fail open-circuit in case of a circuit-failure. They can get hot and must be mounted with some mm clearance above the PCB.
R5 is a Metal-Oxide Varistor ( MOV ) that will limit the voltage to 200..250 Vp.
D1..D4 is a 1 A bridge rectifier with a voltage rating of at least 400 V. The 1N4007 diodes are 1000 V types. These were cheaper than 1N4004 when I bought them.
The circuit around Q2 is a 5 mA constant-current generator.
Q2 is a TO-220 type with a voltage rating of at least 400 V.
Most modern MOS-FETs will work, but some older types may have too large ( >10 µA ) gate-source leakage current ( reducing the value of R6 will fix this - a 180 kΩ resistor will dissipate up to 0.1 W ).
The continuous voltage rating of R6 must be at least 200 V. Most through-hole 0.25 W resistors will work, but check the data-sheet.
The power dissipation in Q2 is 0.63 W at 135 VAC supply voltage, so if you intend to use the circuit for extended periods at high ambient temperature, a clip-on heat-sink is recommended.
The drain current in Q2 is determined by the current required to trig opto-coupler D8.
The IL420 has a maximum trig current of 2 mA at 25 °C and a temperature coefficient of 14 µA/°C resulting in a maximum trig current of 3 mA at 85 °C.
Adding another 30% for LED aging result in a minimum trig current of 4 mA.

Some manufacturers ( of opto-couplers ) recommend the use of zero-crossing opto-couplers for driving relays and some manufacturers ( of relays ) advise against it.
I did the calculation for a zero-crossing opto-coupler ( IL410 ) and the required trig current is 8 mA.
This also requires the addition of power-factor correction ( a resistor and a capacitor in series across the relay-coil ) for the relay. The power-factor of the relay is around 0.5. With 8.2 kΩ and 150 nF, the power-factor is around 0.9 and power dissipation in the resistor is 0.2 W. For unity power-factor, the capacitor must be 250 nF and power dissipation in the resistor is almost 0.5 W.
Power factor correction is not required when using a random-phase opto-coupler.
R11 is a MOV to protect the opto-coupler when the relay is released.

D6 and D7 are voltage regulators for the circuit.
When the relay is open, the voltage across them is 10 V and when the relay is closed, Q3, R10 and D8 will reduce the voltage across D6 to 2 V.
This in turn applies hysteresis to the timer around U1C and the input voltage detector, Q4, R15, R16, D9 and D10. These 5 components are chosen so the temperature coefficient is close to zero from 0 to 85 °C.
Together with the components in the input voltage detector, C4, R14 and U1B will detect a missing AC voltage and reset the circuit. With the component values shown, one missing half-period is detected.
This may be too fast for some applications. Increase C4 or R14 for less sensitivity.
U1 must be a LM193, LM293 or LM393. It is one of the few comparators that will work with its output above the supply ( actually it will work with the output and one of the inputs above the supply ).
Some of the LM393 I have will output a low level until the supply reaches 3 V. C2 prevent the relay from turning on during this period.

PS37 voltage and current waveforms.
Fig.4: PS37 voltage and current waveforms. Ground for the simulation is D7-anode.
Stimulus is a 115 V sine starting at 0 ms, skipping 1 half-period at 360 ms and turning off at 500 ms.

Red: Voltage on C1 plus.
Blue: Voltage on D10 anode. This is scaled by 1/100 on the plot.
Magenta: Voltage on Q4 collector.
Lime: Voltage on U1 pin 6.
Cyan: Voltage on U1 pins 2 and 5.
Black: Current through D8 LED.

Component selection.

The circuit is tested with the components in the parts list ( in the design files download later ), but some components may need to be changed to fit your application.
As this circuit is not optimized for volume production, the suggestions below are conservatively rated.


The fuse used for this application should be time-lag 250 V type.
The rating should be sufficient for the application.
There is no need to use a fuse with a rating beyond the current you use.
This is often done due to transformer inrush current, but is not necessary when using a soft-starter.

You must select a fuse with a sufficient breaking capacity for the application.
The most common breaking capacities for time-lag fuses are 35 A and 1500 A.
The breaking capacity is calculated by dividing the mains voltage by the transformer primary resistance ( see note 1 ).
For example for a transformer with 2 115 VAC primaries ( each 3 Ω ) on 230 VAC:
Breaking capacity ≥ 230 / 6 = 38 A
For the same transformer on 115 VAC:
Breaking capacity ≥ 115 / 1.5 = 77 A
For low volume designs I normally use mains-fuses with 1500 A breaking capacity ( I hate exploding glass fuses ).


The PCB will accept a number of different fuse-clip types. You can find the drill pattern in the PCB component lay-out drawing.

Littelfuse recommends that fuse-holders should be operated at no more than 60% of their rated current.


The PCB will accept a number of different ICL types. You can find the drill pattern in the PCB component lay-out drawing.

The ICLs MUST have a continuous current rating of at least double the fuse rating.
Fuses can carry their rated current continuously. You will need to draw close to double this value to blow the fuse.

The maximum energy that can be dissipated in the ICL is the same as the energy stored in the capacitors and the energy due to the transformer's inrush current.
For example for a supply with a 225 VA transformer, ±35 V nominal output voltage and 2 10 mF capacitors:

Energy due to charging the capacitors:
Nominal mains voltage: 230 VAC
Maximum mains voltage: 265 VAC
Nominal voltage on capacitor: 35 V
Maximum voltage on capacitor: 35 / 230 * 265 = 40 V
Energy stored in one capacitor: 0.5 * 1.0e-2 * 402 = 8 J
This is calculated for each output and the results added ( 16 J for this example ).

Energy due to transformer inrush current ( for a large toroid transformer ):
Energy = 30 * steady state current * mains voltage / ( 2 * mains frequency )
30 is the factor that the inrush current can exceed the transformer's nominal full-load current.
Steady state current is the current the transformer will draw at its rated load, in this case 225 / 230 = 1 A.
Dividing by two times the mains frequency means that this can happen for a half mains period.
This is 68 J for this example, giving a minimum ICL energy rating of 84 J.
You will need a 22 mm ICL or two 15 mm ICLs in series for this.
If you have any doubt about the numbers used for ICL calculation - use bigger ICLs.

There is a spread-sheet with ICL calculations in the design files download.

Resistors R3 and R4

R3 and R4 must be safety resistors.
I have tested the circuit with 2 different types:
Yageo FRM25 are rated at 0.25 W and have no voltage specification ( and are out of production, but it should be possible to find a substitute ).
Vishay NFR25H are rated at 0.5 W and have a continuous voltage rating of 350 V and a peak voltage rating of 1 kV.
I will prefer the NRF25H in this application due to their higher peak voltage rating.

Jumper wires

In some cases you will want to bypass one or more components on the PCB ( in case you do not need the voltage selector, want an external fuse, etc. ).

Table 1: Maximum wire diameter that will fit in the footprints.
ComponentMaximum wire diameter
I11.5 mm
J21.3 mm
R1, R21.5 mm

Table 2: Solid wire current rating in air ( ref [2] ).
AWGWire diameterWire cross sectionMaximum current
240.51 mm0.20 mm23.5 A
230.57 mm0.26 mm24.7 A
220.64 mm0.32 mm27 A
210.72 mm0.41 mm29 A
200.81 mm0.41 mm211 A
190.91 mm0.65 mm214 A
181.0 mm0.82 mm216 A
171.2 mm1.0 mm219 A
161.3 mm1.3 mm222 A
151.5 mm1.7 mm228 A
141.6 mm2.1 mm232 A

Note 1 Some have suggested that a fuse with a breaking capacity that exceed the installation breaker is sufficient. It is not.
I tested it on a mains outlet with a 10 A breaker and 0.25 Ω source impedance by connecting a fuse directly across.
A 1 A fuse with a 1500 A breaking capacity blows with no visible damage to the fuse.
A 1 A fuse with a 35 A breaking capacity blows and trips the 10 A breaker. The end caps and some dust was all that was left of the fuse.
This kind of testing is done in a closed box with proper mains wiring ( no test leads or clips ).


PS37 PCB photo.
Fig.5: Photo of mounted PCB.

Board specification:
Maximum continuous current at 20 °C temperature rise: 11 A
Fusing current at 20 °C ambient temperature: 80 A for 1 s
Trace to trace clearance on "primary" ( fig.2 ) side: 3 mm
Clearance to mounting hole pads and board edges: 6 mm
Board size: 79.4 * 99.1 mm

Download PS37D design files.

I have boards available for this project. See the PCBs page.

Known Issues / updates.

No known issues.


[1] Inrush current limiters for power supplies.
[2] Saturn PCB Toolkit.

Copyright and disclaimer.

Copyright Notice.
This web-page, including but not limited to all text, drawings and photos, is the intellectual property of Poul Petersen, and is Copyright ©.
Reproduction or re-publication by any means whatsoever is strictly prohibited under International Copyright laws.
The author grants the reader the right to use this information for personal use only.
Any commercial use is prohibited without express written authorization from Poul Petersen.

The information is provided on an "as-is" basis and is believed to be correct, however any use of the information is your own responsibility.

This web-site may contain links to web-sites outside Poul Petersen domain ( www.poulpetersen.dk ).
Poul Petersen has no control over and assumes no responsibility for the content of any web-site outside Poul Petersen own domain.

Poul Petersen does not use cookies to "enhance your experience" on this website.

Poul Petersen, C/Faya 14, 35120 Arguineguín, Las Palmas, Spain.
Poul Petersen DIY index

E-mail: diy@poulpetersen.dk

Copyright © Poul Petersen 2019. Last update: 20190601. Valid HTML!